Information processing apparatus, program, and information processing method

Patent metadata

Field	Value
Publication number	US-2016196505-A1
Application number	US-201514861182-A
Country	US
Kind code	A1
Filing date	Sep 22, 2015
Priority date	Sep 22, 2014
Publication date	Jul 7, 2016
Grant date	—

Abstract

Various embodiments train a prediction model for predicting a label to be allocated to a prediction target explanatory variable set. In one embodiment, one or more sets of training data are acquired. Each of the one or more sets of training data includes at least one set of explanatory variables and a label allocated to the at least one explanatory variable set. A plurality of explanatory variable subsets is extracted from the at least one set of explanatory variables. A prediction model is trained utilizing the training data. The plurality of explanatory variable subsets is reflected on a label predicted by the prediction model to be allocated to a prediction target explanatory variable set with each of the plurality of explanatory variable subsets weighted respectively.

First claim

What is claimed is:

1 . A method for training a prediction model for predicting a label to be allocated to a prediction target explanatory variable set, the method comprising:
acquiring one or more sets of training data, each of the one or more sets of training data comprising at least one set of explanatory variables and a label allocated to the at least one explanatory variable set; extracting a plurality of explanatory variable subsets from the at least one set of explanatory variables; and training a prediction model utilizing the training data, where the plurality of explanatory variable subsets is reflected on a label predicted by the prediction model to be allocated to a prediction target explanatory variable set with each of the plurality of explanatory variable subsets weighted respectively.

2 . The method according to claim 1 , wherein training the prediction further comprises:
allocating a different weight coefficient to each of the plurality of explanatory variable subsets.

3 . The method according to claim 2 , further comprising:
generating a feature vector, concerning each of the plurality of explanatory variable subsets, comprising a plurality of feature values, wherein training the prediction model further comprises:
utilizing a regression vector comprising a plurality of regression coefficients respectively corresponding to the plurality of feature values of the feature vector and the weight coefficient of each of plurality of explanatory variable subsets.

4 . The method according to claim 3 , wherein training the prediction model further comprises:
executing Bayesian inference using prior distributions of the regression vector, the weight coefficients, the training data; and outputting a posterior probability distribution of the regression vector and the weight coefficients as a training result.

5 . The method according to claim 4 , wherein training the prediction model further comprises:
utilizing an objective function to be minimized for training the prediction model, the objective function comprising a weighted sum of terms indicating errors between labels predicted for the plurality of explanatory variable subsets based on the feature vector and the regression vector, and the label allocated to the at least one explanatory variable set.

6 . The method according to claim 4 , wherein training the prediction model further comprises:
utilizing, as prior distributions, the output posterior probability distributions of the regression vector and the weight coefficients; and outputting posterior probability distributions of the regression vector and the weight coefficients for additional training data.

7 . The method according to claim 1 , wherein each of the one or more sets of training data is a time-series data set observed over time, and
wherein the extracting further comprises:
extracting, as the plurality of explanatory variable subsets, a plurality of data sequences continuous in time series.

8 . The method according to claim 7 , wherein the plurality of data sequences comprises a set of values of a plurality of feature values in a plurality of sections.

9 . The method according to claim 7 , wherein the plurality of data sequences partially overlapping one another in a time series.

10 . The method according to claim 1 , wherein the acquiring further comprises:
acquiring a prediction target data set comprising a prediction target explanatory variable set, and wherein method further comprises:
predicting a label corresponding to the prediction target explanatory variable set based on the prediction model.

11 . The method according to claim 10 , wherein the training further comprises:
setting, as additional training data, the prediction target data set; and further training the prediction model base on the prediction target data set.

12 . An information processing apparatus for training a prediction model for predicting a label to be allocated to a prediction target explanatory variable set, the information processing apparatus comprising:
a memory; a processor communicatively coupled to the memory; an acquiring unit to acquire one or more sets of training data, each of the one or more sets of training data comprising at least one set of explanatory variables and a label allocated to the at least one explanatory variable set; an extracting unit to extract a plurality of explanatory variable subsets from the at least one set of explanatory variables; and a training processing unit to train a prediction model, where the prediction model is trained utilizing the training data where the plurality of explanatory variable subsets is reflected on a label predicted by the prediction model to be allocated to a prediction target explanatory variable set with each of the plurality of explanatory variable subsets weighted respectively.

13 . The information processing apparatus according to claim 12 , wherein the acquiring unit is further to:
acquire a prediction target data set comprising a prediction target explanatory variable set, and wherein the information processing apparatus further comprises:
a predicting unit to predict a label corresponding to the prediction target explanatory variable set based on the prediction model.

14 . The information processing apparatus according to claim 12 , wherein the training processing unit trains the prediction model by allocating a different weight coefficient to each of the plurality of explanatory variable subsets.

15 . The information processing apparatus according to claim 14 , further comprising a feature vector generating unit to generate a feature vector comprising a plurality of feature values, concerning each of the plurality of explanatory variable subsets, wherein
the training processing unit trains the prediction model by utilizing a regression vector comprising a plurality of regression coefficients respectively corresponding to the plurality of feature values of the feature vector and the weight coefficient of each of plurality of explanatory variable subsets.

16 . A program product for causing a computer to training a prediction model for predicting a label to be allocated to a prediction target explanatory variable set, the program product, when executed, causes the computer to perform a method comprising:
acquiring one or more sets of training data, each of the one or more sets of training data comprising at least one set of explanatory variables and a label allocated to the at least one explanatory variable set; extracting a plurality of explanatory variable subsets from the at least one set of explanatory variables; and training a prediction model utilizing the training data, where the plurality of explanatory variable subsets is reflected on a label predicted by the prediction model to be allocated to a prediction target explanatory variable set with each of the plurality of explanatory variable subsets weighted respectively.

17 . The program product according to claim 16 , where training the prediction further comprises:
allocating a different weight coefficient to each of the plurality of explanatory variable subsets.

18 . The program product according to claim 17 , wherein the method further comprises:
generating a feature vector, concerning each of the plurality of explanatory variable subsets, comprising a plurality of feature values, wherein training the prediction model further comprises:
utilizing a regression vector comprising a plurality of regression coefficients respectively correspondin

Assignees

• IBM

Inventors

• Katsuki Takayuki
• Shinohara Yuma

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N7/005

  Physics · mapped topic

• G06N99/005 Primary

  Physics · mapped topic

• G06N20/00 Primary

  Machine learning · CPC title